Field of the Invention
The present invention relates to techniques for controlling imaging processing.
Description of the Related Art
Hitherto, there have been proposed techniques for determining an object to be imaged (hereinafter referred to as an “imaging object”), e.g., a suspicious person in a building or a room, in which an image capturing apparatus is placed, and for tracking and imaging the object with the image capturing apparatus.
Regarding such techniques, e.g., Japanese Patent Application Laid-Open No. 6-325180 discusses a method in which a person such as an observer specifies a person to be tracked (hereinafter referred to as a tracking object) and in which a plurality of image capturing apparatuses performs tracking of the specified person in cooperation with one another. Further, Japanese Patent Application Laid-Open No. 2007-67963 discusses a method for sequentially assigning, in the case of a changing relationship between the number of image capturing apparatuses and imaging objects, the image capturing apparatuses to the imaging objects.
Generally, in biometric authentication, the success or failure of authentication is determined by comparison between a similarity, which is calculated by collating biometric information read by a biometric authentication apparatus with a template (preliminarily registered true biometric information) and a preliminarily set threshold.
FIG. 9 illustrates normalized frequency distributions of similarity, which respectively correspond to an impostor template and a client template. The axis of abscissa represents similarity. The axis of ordinate represents frequency. Therefore, the similarity increases to the right, as viewed in FIG. 9. This means that the degree of match between two pieces of biometric information, which are to be compared with each other, increases to the right. In a case where the similarity is equal to or higher than a predetermined threshold, it is determined that the two pieces of biometric information match each other. Otherwise, it is determined that the two pieces of biometric information do not match each other.
Biometric authentication includes a trade-off between the false rejection rate (FRR) and the false acceptance rate (FAR). The FRR is a probability at which an identical person to be verified (hereinafter called a “client”) is falsely rejected as any person other than the client (hereinafter called an “impostor”). The FAR is a probability at which an imposer is falsely accepted as a true client.
The ratio of the FRR to the FAR is determined by setting a threshold. When the threshold is set at a small value, the FRR, at which a true client is falsely rejected, is low, while the FAR, at which an imposer is falsely accepted, is high. Conversely, when the threshold is set at a large value, the FAR, at which an imposer is falsely accepted, is low, while the FRR, at which a true client is falsely rejected, is high. Specifically, in a building and a room the security of each of which is treated as important, it is necessary to improve the security by monitoring a person who enters the building or the room while being falsely accepted at the FAR.
Meanwhile, according to the method discussed in Japanese Patent Application Laid-Open No. 6-325180, an observer determines whether the specified person is a proper tracking object. Accordingly, the observer should be very familiar with users and usages of the building or the room. In this case, the observer bears a heavy load. In addition, the observer might be unable to make a correct decision. Thus, security accuracy cannot be sufficient.
According to the method discussed in Japanese Patent Application Laid-Open No. 2007-67963, the image capturing apparatuses are sequentially assigned to the imaging objects. Thus, even in a case where a person who is being imaged is a proper tracking object, the image capturing apparatus may be assigned to a person other than the proper person. Accordingly, the method discussed in Japanese Patent Application Laid-Open No. 2007-67963 is unsuitable for tracking a specific person. Thus, security accuracy cannot be sufficient.
The level of observation of a person in a building and a room, the security of each of which is treated as important, can be increased by recording a video image of the person imaged by the image capturing apparatus. However, in a case where a video image of a person is always recorded, a huge volume of recorded image data is accumulated. Accordingly, a system is required, which is adapted to automatically record image data only when the image data is needed.